This action might not be possible to undo. Are you sure you want to continue?
Introduction Geomodelling aims at representing as well as possible the reservoir heterogeneities in terms of lithofacies and petrophysical variables. Consequently, the modelling methods and their optimization play an important role for a refined and adequate modelling. When modelling, the issues to tackle are, among others: the level of realism of the facies simulations, the conditioning of simulations to wells data and the spatial behavior (local anisotropy integration for instance). Several simulations methods are available either in geological modelling or in petrophysical modelling. All those methods have their pros and cons. For example, facies variogram-based methods as Sequential indicator Simulation (SIS), Plurigaussian simulation (PGS) are flexible but may not be able to properly represent connectivity in some cases. On the contrary, object-based methods (as Boolean) simulate connected bodies but the data conditioning is not easy. Multiple-points Statistics simulations (MPS), at the edge between pixel-based and object-based methods, may look more promising but needs reliable and consistent prior information for the facies distribution and relationships (geological training image). This paper aims at combining several algorithms in order to benefit from their advantages and therefore optimize the modelling. The following workflow is proposed: first, a process-based algorithm for Meandering Channels simulations is run in order to generate realistic facies simulations. This geological information is then used as a training image for MPS simulations, allowing efficient data conditioning. Finally, locally varying anisotropies are integrated for ensuring a better continuity of porosity simulations in the main direction of each channel. Material and modelling methods The proposed modelling workflow relies on different approaches. Firstly Meandering Channels simulations are run, in order to reproduce a fluvial environment. Then, a geological information (training image) is extracted from one Meandering Channels simulation in order to be used as input in MPS simulations. Afterwards a porosity modelling is carried out, constrained by information retrieved from meandering channels simulation. A local optimization of the spatial parameters is done (local variogram anisotropies). These steps are detailed below and then applied on a synthetic dataset. Meandering Channels simulations are based on a genetic (process-based) algorithm called FLUMY (Rivoirard et al., 2007) that simulate fluvial depositional environment system. Processes such as channel migration, overbank flood, levee breaching, water table can be reproduced during the simulation. Aggradations are setup during time in order to bring changes during the life of the depositional system. Such process-based approach gives realistic bodies and arrangements. The complexity of sand bodies and the contrast arrangements are reproduced, therefore producing geologically realistic simulations. However, although possible in some cases, data conditioning may be tricky with such algorithms. MPS simulations are a simulation method that reproduces geological connected bodies (as channels) contained in a prior “training image” (Strebelle, 2000; Strebelle, 2002) The training image represents the geological context to be simulated. It can be either in 2D or 3D. It is derived from the geological knowledge of the reservoir. Analogs, outcrops, geological conceptual model can be used to build a training image. Geological bodies’ simulations (object-oriented simulations, meandering channel simulations) can also be used. When running variogram-based simulations of porosity, a local optimization of the variogram (and neighborhood) parameters can be done. Local ranges, local anisotropies can be computed for each simulated node using the Local Geostatistics approach (Magneron et al., 2009). These parameters are integrated in the SGS algorithm. Local optimization allows integrating geological heretogeneities
75th EAGE Conference & Exhibition incorporating SPE EUROPEC 2013 London, UK, 10-13 June 2013
It allows to have a more accurate modelling and to reduce uncertainties. 10-13 June 2013 . see Figure 2 below) and for porosity modelling (see Figure 4 and Figure 5). overbank alluvium). Results 1) Meandering Channels simulations Non conditional Meandering Channels simulations have been run using the FLUMY approach. They are used for multiple-points simulations (as training image. Lithotypes Point Bar 10000 Levee Overbank Alluvium Y (m) 5000 0 0 5000 X (m) 10000 15000 Figure 2: 2D Training image extracted from a meandering channels simulation. 75th EAGE Conference & Exhibition incorporating SPE EUROPEC 2013 London.information. Lithotypes Point Bar Levee Overbank Alluvium Figure 1: 3D Meandering Channels simulation. levee. A simple lithological model based on three lithotypes has been used (point bar. For example. 2) Geological information extraction From 3D meandering channels simulations. the variogram main anisotropy direction will change according to the channel orientation. 2D time slices are extracted. UK.
3) Lithofacies multiple-points statistics simulations For illustration purpose. Lithotypes 5000 Point Bar Levee Overbank Alluvium 4000 Y (m) 3000 2000 1000 0 0 1000 2000 3000 4000 X (m) 5000 6000 7000 Figure 3: 2D Multiple-Points facies simulation. But other local parameters (local ranges. used as training image. UK. local sills) could also be used. 75th EAGE Conference & Exhibition incorporating SPE EUROPEC 2013 London. by construction. In that example. only one local parameter has been computed. Such MPS simulations can. 2D simulations are carried out. 4) Porosity simulations with local optimization of geostatistical parameters Channels paths have been extracted from a meandering channel simulation. Then the local azimuth has been computed along the channels through a local variogram cross-validation process. easily be conditioned to honour input wells data. The main point is to reproduce the connectivity of the channels visible in Figure 2. Azimuth rotation 90 70 2000 50 30 Y (m) 10 -10 1000 -30 -50 -70 -90 0 0 1000 X (m) 2000 Figure 4: Locally optimized parameter: azimuth rotation. 10-13 June 2013 .
Noise reduction by M-factorial kriging. 11th International Congress of the Brazilian Geophysical Society. Strebelle S. (2007) Controling the Location of Deposits in Meandering Channelized Reservoir Models. 2000 2000 Y (m) 1000 Y (m) 1000 0 0 1000 X (m) 2000 0 0 1000 X (m) 2000 Figure 5: A porosity simulation (left). The Netherlands. Stanford. Portugal. local geostatistics are carried out. 19-24. pp. a locally azimuth optimized porosity simulation (right). 10-13 June 2013 . thesis. and Geffroy. Petroleum Geostatistics 2007. PhD. Such methods and procedures allow also reducing the uncertainties attached to the modelling. Mathematical Geology 34(1): 1-22. Strebelle S (2002) Conditional simulation of complex geological structures using multiple point geostatistics. Rivoirard. (2000) Sequential simulation drawing structures from training images. 2012). Bourges M. 252 p. Acknowledgements All geostatistical computations are performed using Isatis© software (Geovariances.. To optimize the porosity modelling. Stanford University. Sequential Gaussian simulations (SGS) of porosity have been carried out within the channels. Conclusions Combining various approaches in geomodelling (for facies or for petrophysical properties) can be interesting in order to get a more realistic and accurate reservoir model.Using the local azimuth rotation. Cojan. Bahia. 10-14 September 2007. European Association of Geoscientists & Engineers. (2009). SBGF. References Geovariances (2012). Geological information is then retrieved to run a lithofacies multiple-point simulation and a porosity modelling. F. I. In this paper.. Department of Geological and Environmental Sciences. The local parameter reinforces the porosity continuity along the channels (see Figure 5 below) compared to a classical isotropic simulation. UK. Magneron C. meandering channels simulations allowed to build a geologically realistic model. Jeannée N. in Extended Abstracts Book.Cascais. Isatis technical references.. 75th EAGE Conference & Exhibition incorporating SPE EUROPEC 2013 London. Houten. J.
This action might not be possible to undo. Are you sure you want to continue?
We've moved you to where you read on your other device.
Get the full title to continue listening from where you left off, or restart the preview.